REOH: Runtime Energy Optimization for Heterogeneous Systems

Significant efforts have been devoted to choosing the best configuration of a computing system to run an application energy efficiently. However, available tuning approaches mainly focus on homogeneous systems and are inextensible for heterogeneous systems which include several components (e.g., CPUs, GPUs) with different architectures. This study proposes a holistic tuning approach called REOH, based on probabilistic models to predict the most energy-efficient configuration (i.e., which platform and its setting) of a heterogeneous system for running a given application. Based on the computation and communication patterns from Berkeley dwarfs, we conduct experiments to devise the training data set that covers a wide range of application patterns and characteristics. Validating the REOH approach on heterogeneous systems including CPUs and GPUs shows that the energy consumption by the REOH approach is only one third the energy consumption by the previous (homogeneous) approaches in the best cases and close to the optimal energy consumption by the brute-force approach. Based on the REOH approach, we develop a open-source energy-optimizing runtime framework for selecting an energy efficient configuration of a heterogeneous system for a given application at runtime.